/* Copyright 2017 The Kubernetes Authors. Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. */ package crictl import ( "encoding/json" "fmt" "log" "sort" "strings" "time" "github.com/docker/go-units" godigest "github.com/opencontainers/go-digest" "github.com/sirupsen/logrus" "github.com/urfave/cli" "golang.org/x/net/context" pb "k8s.io/cri-api/pkg/apis/runtime/v1alpha2" ) type containerByCreated []*pb.Container func (a containerByCreated) Len() int { return len(a) } func (a containerByCreated) Swap(i, j int) { a[i], a[j] = a[j], a[i] } func (a containerByCreated) Less(i, j int) bool { return a[i].CreatedAt > a[j].CreatedAt } type createOptions struct { // podID of the container podID string // the config and pod options *runOptions } type runOptions struct { // configPath is path to the config for container configPath string // podConfig is path to the config for sandbox podConfig string // the image pull options *pullOptions } type pullOptions struct { // do not pull the image on container creation dontPull bool // creds is string in the format `USERNAME[:PASSWORD]` for accessing the // registry during image pull creds string // auth is a base64 encoded 'USERNAME[:PASSWORD]' string used for // authentication with a registry when pulling an image auth string } var pullFlags = []cli.Flag{ cli.BoolFlag{ Name: "no-pull", Usage: "Do not pull the image on container creation", }, cli.StringFlag{ Name: "creds", Value: "", Usage: "Use `USERNAME[:PASSWORD]` for accessing the registry", }, cli.StringFlag{ Name: "auth", Value: "", Usage: "Use `AUTH_STRING` for accessing the registry. AUTH_STRING is a base64 encoded 'USERNAME[:PASSWORD]'", }, } var createContainerCommand = cli.Command{ Name: "create", Usage: "Create a new container", ArgsUsage: "POD container-config.[json|yaml] pod-config.[json|yaml]", Flags: pullFlags, Action: func(context *cli.Context) error { if len(context.Args()) != 3 { return cli.ShowSubcommandHelp(context) } runtimeClient, runtimeConn, err := getRuntimeClient(context) if err != nil { return err } defer closeConnection(context, runtimeConn) imageClient, imageConn, err := getImageClient(context) if err != nil { return err } defer closeConnection(context, imageConn) opts := createOptions{ podID: context.Args().Get(0), runOptions: &runOptions{ configPath: context.Args().Get(1), podConfig: context.Args().Get(2), pullOptions: &pullOptions{ dontPull: context.Bool("no-pull"), creds: context.String("creds"), auth: context.String("auth"), }, }, } ctrID, err := CreateContainer(imageClient, runtimeClient, opts) if err != nil { return fmt.Errorf("Creating container failed: %v", err) } fmt.Println(ctrID) return nil }, } var startContainerCommand = cli.Command{ Name: "start", Usage: "Start one or more created containers", ArgsUsage: "CONTAINER-ID [CONTAINER-ID...]", Action: func(context *cli.Context) error { if context.NArg() == 0 { return cli.ShowSubcommandHelp(context) } runtimeClient, runtimeConn, err := getRuntimeClient(context) if err != nil { return err } defer closeConnection(context, runtimeConn) for i := 0; i < context.NArg(); i++ { containerID := context.Args().Get(i) err := StartContainer(runtimeClient, containerID) if err != nil { return fmt.Errorf("Starting the container %q failed: %v", containerID, err) } } return nil }, } var updateContainerCommand = cli.Command{ Name: "update", Usage: "Update one or more running containers", ArgsUsage: "CONTAINER-ID [CONTAINER-ID...]", Flags: []cli.Flag{ cli.Int64Flag{ Name: "cpu-period", Usage: "CPU CFS period to be used for hardcapping (in usecs). 0 to use system default", }, cli.Int64Flag{ Name: "cpu-quota", Usage: "CPU CFS hardcap limit (in usecs). Allowed cpu time in a given period", }, cli.Int64Flag{ Name: "cpu-share", Usage: "CPU shares (relative weight vs. other containers)", }, cli.Int64Flag{ Name: "memory", Usage: "Memory limit (in bytes)", }, cli.StringFlag{ Name: "cpuset-cpus", Usage: "CPU(s) to use", }, cli.StringFlag{ Name: "cpuset-mems", Usage: "Memory node(s) to use", }, }, Action: func(context *cli.Context) error { if context.NArg() == 0 { return cli.ShowSubcommandHelp(context) } runtimeClient, runtimeConn, err := getRuntimeClient(context) if err != nil { return err } defer closeConnection(context, runtimeConn) options := updateOptions{ CPUPeriod: context.Int64("cpu-period"), CPUQuota: context.Int64("cpu-quota"), CPUShares: context.Int64("cpu-share"), CpusetCpus: context.String("cpuset-cpus"), CpusetMems: context.String("cpuset-mems"), MemoryLimitInBytes: context.Int64("memory"), } for i := 0; i < context.NArg(); i++ { containerID := context.Args().Get(i) err := UpdateContainerResources(runtimeClient, containerID, options) if err != nil { return fmt.Errorf("Updating container resources for %q failed: %v", containerID, err) } } return nil }, } var stopContainerCommand = cli.Command{ Name: "stop", Usage: "Stop one or more running containers", ArgsUsage: "CONTAINER-ID [CONTAINER-ID...]", SkipArgReorder: true, UseShortOptionHandling: true, Flags: []cli.Flag{ cli.Int64Flag{ Name: "timeout, t", Value: 10, Usage: "Seconds to wait to kill the container after a graceful stop is requested", }, }, Action: func(context *cli.Context) error { if context.NArg() == 0 { return cli.ShowSubcommandHelp(context) } runtimeClient, runtimeConn, err := getRuntimeClient(context) if err != nil { return err } defer closeConnection(context, runtimeConn) for i := 0; i < context.NArg(); i++ { containerID := context.Args().Get(i) err := StopContainer(runtimeClient, containerID, context.Int64("timeout")) if err != nil { return fmt.Errorf("Stopping the container %q failed: %v", containerID, err) } } return nil }, } var removeContainerCommand = cli.Command{ Name: "rm", Usage: "Remove one or more containers", ArgsUsage: "CONTAINER-ID [CONTAINER-ID...]", SkipArgReorder: true, UseShortOptionHandling: true, Flags: []cli.Flag{ cli.BoolFlag{ Name: "force, f", Usage: "Force removal of the container, disregarding if running", }, cli.BoolFlag{ Name: "all, a", Usage: "Remove all containers", }, }, Action: func(ctx *cli.Context) error { runtimeClient, runtimeConn, err := getRuntimeClient(ctx) if err != nil { return err } defer closeConnection(ctx, runtimeConn) ids := ctx.Args() if ctx.Bool("all") { r, err := runtimeClient.ListContainers(context.Background(), &pb.ListContainersRequest{}) if err != nil { return err } ids = nil for _, ctr := range r.GetContainers() { ids = append(ids, ctr.GetId()) } } if len(ids) == 0 { return cli.ShowSubcommandHelp(ctx) } errored := false for _, id := range ids { resp, err := runtimeClient.ContainerStatus(context.Background(), &pb.ContainerStatusRequest{ContainerId: id}) if err != nil { logrus.Error(err) errored = true continue } if resp.GetStatus().GetState() == pb.ContainerState_CONTAINER_RUNNING { if ctx.Bool("force") { if err := RemoveContainer(runtimeClient, id); err != nil { logrus.Errorf("stopping the container %q failed: %v", id, err) errored = true continue } continue } else { logrus.Errorf("container %q is running, please stop it first", id) errored = true continue } } err = RemoveContainer(runtimeClient, id) if err != nil { logrus.Errorf("removing container %q failed: %v", id, err) errored = true continue } } if errored { return fmt.Errorf("unable to remove container(s)") } return nil }, } var containerStatusCommand = cli.Command{ Name: "inspect", Usage: "Display the status of one or more containers", ArgsUsage: "CONTAINER-ID [CONTAINER-ID...]", Flags: []cli.Flag{ cli.StringFlag{ Name: "output, o", Usage: "Output format, One of: json|yaml|table", }, cli.BoolFlag{ Name: "quiet, q", Usage: "Do not show verbose information", }, }, Action: func(context *cli.Context) error { if context.NArg() == 0 { return cli.ShowSubcommandHelp(context) } runtimeClient, runtimeConn, err := getRuntimeClient(context) if err != nil { return err } defer closeConnection(context, runtimeConn) for i := 0; i < context.NArg(); i++ { containerID := context.Args().Get(i) err := ContainerStatus(runtimeClient, containerID, context.String("output"), context.Bool("quiet")) if err != nil { return fmt.Errorf("Getting the status of the container %q failed: %v", containerID, err) } } return nil }, } var listContainersCommand = cli.Command{ Name: "ps", Usage: "List containers", SkipArgReorder: true, UseShortOptionHandling: true, Flags: []cli.Flag{ cli.BoolFlag{ Name: "verbose, v", Usage: "Show verbose information for containers", }, cli.StringFlag{ Name: "id", Value: "", Usage: "Filter by container id", }, cli.StringFlag{ Name: "name", Value: "", Usage: "filter by container name regular expression pattern", }, cli.StringFlag{ Name: "pod, p", Value: "", Usage: "Filter by pod id", }, cli.StringFlag{ Name: "image", Value: "", Usage: "Filter by container image", }, cli.StringFlag{ Name: "state", Value: "", Usage: "Filter by container state", }, cli.StringSliceFlag{ Name: "label", Usage: "Filter by key=value label", }, cli.BoolFlag{ Name: "quiet, q", Usage: "Only display container IDs", }, cli.StringFlag{ Name: "output, o", Usage: "Output format, One of: json|yaml|table", }, cli.BoolFlag{ Name: "all, a", Usage: "Show all containers", }, cli.BoolFlag{ Name: "latest, l", Usage: "Show the most recently created container (includes all states)", }, cli.IntFlag{ Name: "last, n", Usage: "Show last n recently created containers (includes all states). Set 0 for unlimited.", }, cli.BoolFlag{ Name: "no-trunc", Usage: "Show output without truncating the ID", }, }, Action: func(context *cli.Context) error { runtimeClient, runtimeConn, err := getRuntimeClient(context) if err != nil { return err } defer closeConnection(context, runtimeConn) imageClient, imageConn, err := getImageClient(context) if err != nil { return err } defer closeConnection(context, imageConn) opts := listOptions{ id: context.String("id"), podID: context.String("pod"), state: context.String("state"), verbose: context.Bool("verbose"), quiet: context.Bool("quiet"), output: context.String("output"), all: context.Bool("all"), nameRegexp: context.String("name"), latest: context.Bool("latest"), last: context.Int("last"), noTrunc: context.Bool("no-trunc"), image: context.String("image"), } opts.labels, err = parseLabelStringSlice(context.StringSlice("label")) if err != nil { return err } if err = ListContainers(runtimeClient, imageClient, opts); err != nil { return fmt.Errorf("listing containers failed: %v", err) } return nil }, } var runContainerCommand = cli.Command{ Name: "run", Usage: "Run a new container inside a sandbox", ArgsUsage: "container-config.[json|yaml] pod-config.[json|yaml]", Flags: append(pullFlags, cli.StringFlag{ Name: "runtime, r", Usage: "Runtime handler to use. Available options are defined by the container runtime.", }), Action: func(context *cli.Context) error { if len(context.Args()) != 2 { return cli.ShowSubcommandHelp(context) } runtimeClient, runtimeConn, err := getRuntimeClient(context) if err != nil { return err } defer closeConnection(context, runtimeConn) imageClient, imageConn, err := getImageClient(context) if err != nil { return err } defer closeConnection(context, imageConn) opts := runOptions{ configPath: context.Args().Get(0), podConfig: context.Args().Get(1), pullOptions: &pullOptions{ dontPull: context.Bool("no-pull"), creds: context.String("creds"), auth: context.String("auth"), }, } err = RunContainer(imageClient, runtimeClient, opts, context.String("runtime")) if err != nil { return fmt.Errorf("Running container failed: %v", err) } return nil }, } // RunContainer starts a container in the provided sandbox func RunContainer( iClient pb.ImageServiceClient, rClient pb.RuntimeServiceClient, opts runOptions, runtime string, ) error { // Create the pod podSandboxConfig, err := loadPodSandboxConfig(opts.podConfig) if err != nil { return fmt.Errorf("load podSandboxConfig failed: %v", err) } podID, err := RunPodSandbox(rClient, podSandboxConfig, runtime) if err != nil { return fmt.Errorf("run pod sandbox failed: %v", err) } // Create the container containerOptions := createOptions{podID, &opts} ctrID, err := CreateContainer(iClient, rClient, containerOptions) if err != nil { return fmt.Errorf("creating container failed: %v", err) } // Start the container err = StartContainer(rClient, ctrID) if err != nil { return fmt.Errorf("starting the container %q failed: %v", ctrID, err) } return nil } // CreateContainer sends a CreateContainerRequest to the server, and parses // the returned CreateContainerResponse. func CreateContainer( iClient pb.ImageServiceClient, rClient pb.RuntimeServiceClient, opts createOptions, ) (string, error) { config, err := loadContainerConfig(opts.configPath) if err != nil { return "", err } var podConfig *pb.PodSandboxConfig if opts.podConfig != "" { podConfig, err = loadPodSandboxConfig(opts.podConfig) if err != nil { return "", err } } if !opts.dontPull { auth, err := getAuth(opts.creds, opts.auth) if err != nil { return "", err } // Try to pull the image before container creation image := config.GetImage().GetImage() if _, err := PullImage(iClient, image, auth); err != nil { return "", err } } request := &pb.CreateContainerRequest{ PodSandboxId: opts.podID, Config: config, SandboxConfig: podConfig, } logrus.Debugf("CreateContainerRequest: %v", request) r, err := rClient.CreateContainer(context.Background(), request) logrus.Debugf("CreateContainerResponse: %v", r) if err != nil { return "", err } return r.ContainerId, nil } // StartContainer sends a StartContainerRequest to the server, and parses // the returned StartContainerResponse. func StartContainer(client pb.RuntimeServiceClient, ID string) error { if ID == "" { return fmt.Errorf("ID cannot be empty") } request := &pb.StartContainerRequest{ ContainerId: ID, } logrus.Debugf("StartContainerRequest: %v", request) r, err := client.StartContainer(context.Background(), request) logrus.Debugf("StartContainerResponse: %v", r) if err != nil { return err } fmt.Println(ID) return nil } type updateOptions struct { // CPU CFS (Completely Fair Scheduler) period. Default: 0 (not specified). CPUPeriod int64 // CPU CFS (Completely Fair Scheduler) quota. Default: 0 (not specified). CPUQuota int64 // CPU shares (relative weight vs. other containers). Default: 0 (not specified). CPUShares int64 // Memory limit in bytes. Default: 0 (not specified). MemoryLimitInBytes int64 // OOMScoreAdj adjusts the oom-killer score. Default: 0 (not specified). OomScoreAdj int64 // CpusetCpus constrains the allowed set of logical CPUs. Default: "" (not specified). CpusetCpus string // CpusetMems constrains the allowed set of memory nodes. Default: "" (not specified). CpusetMems string } // UpdateContainerResources sends an UpdateContainerResourcesRequest to the server, and parses // the returned UpdateContainerResourcesResponse. func UpdateContainerResources(client pb.RuntimeServiceClient, ID string, opts updateOptions) error { if ID == "" { return fmt.Errorf("ID cannot be empty") } request := &pb.UpdateContainerResourcesRequest{ ContainerId: ID, Linux: &pb.LinuxContainerResources{ CpuPeriod: opts.CPUPeriod, CpuQuota: opts.CPUQuota, CpuShares: opts.CPUShares, CpusetCpus: opts.CpusetCpus, CpusetMems: opts.CpusetMems, MemoryLimitInBytes: opts.MemoryLimitInBytes, OomScoreAdj: opts.OomScoreAdj, }, } logrus.Debugf("UpdateContainerResourcesRequest: %v", request) r, err := client.UpdateContainerResources(context.Background(), request) logrus.Debugf("UpdateContainerResourcesResponse: %v", r) if err != nil { return err } fmt.Println(ID) return nil } // StopContainer sends a StopContainerRequest to the server, and parses // the returned StopContainerResponse. func StopContainer(client pb.RuntimeServiceClient, ID string, timeout int64) error { if ID == "" { return fmt.Errorf("ID cannot be empty") } request := &pb.StopContainerRequest{ ContainerId: ID, Timeout: timeout, } logrus.Debugf("StopContainerRequest: %v", request) r, err := client.StopContainer(context.Background(), request) logrus.Debugf("StopContainerResponse: %v", r) if err != nil { return err } fmt.Println(ID) return nil } // RemoveContainer sends a RemoveContainerRequest to the server, and parses // the returned RemoveContainerResponse. func RemoveContainer(client pb.RuntimeServiceClient, ID string) error { if ID == "" { return fmt.Errorf("ID cannot be empty") } request := &pb.RemoveContainerRequest{ ContainerId: ID, } logrus.Debugf("RemoveContainerRequest: %v", request) r, err := client.RemoveContainer(context.Background(), request) logrus.Debugf("RemoveContainerResponse: %v", r) if err != nil { return err } fmt.Println(ID) return nil } // marshalContainerStatus converts container status into string and converts // the timestamps into readable format. func marshalContainerStatus(cs *pb.ContainerStatus) (string, error) { statusStr, err := protobufObjectToJSON(cs) if err != nil { return "", err } jsonMap := make(map[string]interface{}) err = json.Unmarshal([]byte(statusStr), &jsonMap) if err != nil { return "", err } jsonMap["createdAt"] = time.Unix(0, cs.CreatedAt).Format(time.RFC3339Nano) jsonMap["startedAt"] = time.Unix(0, cs.StartedAt).Format(time.RFC3339Nano) jsonMap["finishedAt"] = time.Unix(0, cs.FinishedAt).Format(time.RFC3339Nano) return marshalMapInOrder(jsonMap, *cs) } // ContainerStatus sends a ContainerStatusRequest to the server, and parses // the returned ContainerStatusResponse. func ContainerStatus(client pb.RuntimeServiceClient, ID, output string, quiet bool) error { verbose := !(quiet) if output == "" { // default to json output output = "json" } if ID == "" { return fmt.Errorf("ID cannot be empty") } request := &pb.ContainerStatusRequest{ ContainerId: ID, Verbose: verbose, } logrus.Debugf("ContainerStatusRequest: %v", request) r, err := client.ContainerStatus(context.Background(), request) logrus.Debugf("ContainerStatusResponse: %v", r) if err != nil { return err } status, err := marshalContainerStatus(r.Status) if err != nil { return err } switch output { case "json", "yaml": return outputStatusInfo(status, r.Info, output) case "table": // table output is after this switch block default: return fmt.Errorf("output option cannot be %s", output) } // output in table format fmt.Printf("ID: %s\n", r.Status.Id) if r.Status.Metadata != nil { if r.Status.Metadata.Name != "" { fmt.Printf("Name: %s\n", r.Status.Metadata.Name) } if r.Status.Metadata.Attempt != 0 { fmt.Printf("Attempt: %v\n", r.Status.Metadata.Attempt) } } fmt.Printf("State: %s\n", r.Status.State) ctm := time.Unix(0, r.Status.CreatedAt) fmt.Printf("Created: %v\n", units.HumanDuration(time.Now().UTC().Sub(ctm))+" ago") if r.Status.State != pb.ContainerState_CONTAINER_CREATED { stm := time.Unix(0, r.Status.StartedAt) fmt.Printf("Started: %v\n", units.HumanDuration(time.Now().UTC().Sub(stm))+" ago") } if r.Status.State == pb.ContainerState_CONTAINER_EXITED { if r.Status.FinishedAt > 0 { ftm := time.Unix(0, r.Status.FinishedAt) fmt.Printf("Finished: %v\n", units.HumanDuration(time.Now().UTC().Sub(ftm))+" ago") } fmt.Printf("Exit Code: %v\n", r.Status.ExitCode) } if r.Status.Labels != nil { fmt.Println("Labels:") for _, k := range getSortedKeys(r.Status.Labels) { fmt.Printf("\t%s -> %s\n", k, r.Status.Labels[k]) } } if r.Status.Annotations != nil { fmt.Println("Annotations:") for _, k := range getSortedKeys(r.Status.Annotations) { fmt.Printf("\t%s -> %s\n", k, r.Status.Annotations[k]) } } if verbose { fmt.Printf("Info: %v\n", r.GetInfo()) } return nil } // ListContainers sends a ListContainerRequest to the server, and parses // the returned ListContainerResponse. func ListContainers(runtimeClient pb.RuntimeServiceClient, imageClient pb.ImageServiceClient, opts listOptions) error { filter := &pb.ContainerFilter{} if opts.id != "" { filter.Id = opts.id } if opts.podID != "" { filter.PodSandboxId = opts.podID } st := &pb.ContainerStateValue{} if !opts.all && opts.state == "" { st.State = pb.ContainerState_CONTAINER_RUNNING filter.State = st } if opts.state != "" { st.State = pb.ContainerState_CONTAINER_UNKNOWN switch strings.ToLower(opts.state) { case "created": st.State = pb.ContainerState_CONTAINER_CREATED filter.State = st case "running": st.State = pb.ContainerState_CONTAINER_RUNNING filter.State = st case "exited": st.State = pb.ContainerState_CONTAINER_EXITED filter.State = st case "unknown": st.State = pb.ContainerState_CONTAINER_UNKNOWN filter.State = st default: log.Fatalf("--state should be one of created, running, exited or unknown") } } if opts.latest || opts.last > 0 { // Do not filter by state if latest/last is specified. filter.State = nil } if opts.labels != nil { filter.LabelSelector = opts.labels } request := &pb.ListContainersRequest{ Filter: filter, } logrus.Debugf("ListContainerRequest: %v", request) r, err := runtimeClient.ListContainers(context.Background(), request) logrus.Debugf("ListContainerResponse: %v", r) if err != nil { return err } r.Containers = getContainersList(r.GetContainers(), opts) switch opts.output { case "json": return outputProtobufObjAsJSON(r) case "yaml": return outputProtobufObjAsYAML(r) case "table", "": // continue; output will be generated after the switch block ends. default: return fmt.Errorf("unsupported output format %q", opts.output) } display := newTableDisplay(20, 1, 3, ' ', 0) if !opts.verbose && !opts.quiet { display.AddRow([]string{columnContainer, columnImage, columnCreated, columnState, columnName, columnAttempt, columnPodID}) } for _, c := range r.Containers { // Filter by pod name/namespace regular expressions. if !matchesRegex(opts.nameRegexp, c.Metadata.Name) { continue } if match, err := matchesImage(imageClient, opts.image, c.GetImage().GetImage()); err != nil { return fmt.Errorf("failed to check image match %v", err) } else if !match { continue } if opts.quiet { fmt.Printf("%s\n", c.Id) continue } createdAt := time.Unix(0, c.CreatedAt) ctm := units.HumanDuration(time.Now().UTC().Sub(createdAt)) + " ago" if !opts.verbose { id := c.Id image := c.Image.Image if !opts.noTrunc { id = getTruncatedID(id, "") // Now c.Image.Image is imageID in kubelet. if digest, err := godigest.Parse(image); err == nil { image = getTruncatedID(digest.String(), string(digest.Algorithm())+":") } } PodID := getTruncatedID(c.PodSandboxId, "") display.AddRow([]string{id, image, ctm, convertContainerState(c.State), c.Metadata.Name, fmt.Sprintf("%d", c.Metadata.Attempt), PodID}) continue } fmt.Printf("ID: %s\n", c.Id) fmt.Printf("PodID: %s\n", c.PodSandboxId) if c.Metadata != nil { if c.Metadata.Name != "" { fmt.Printf("Name: %s\n", c.Metadata.Name) } fmt.Printf("Attempt: %v\n", c.Metadata.Attempt) } fmt.Printf("State: %s\n", convertContainerState(c.State)) if c.Image != nil { fmt.Printf("Image: %s\n", c.Image.Image) } fmt.Printf("Created: %v\n", ctm) if c.Labels != nil { fmt.Println("Labels:") for _, k := range getSortedKeys(c.Labels) { fmt.Printf("\t%s -> %s\n", k, c.Labels[k]) } } if c.Annotations != nil { fmt.Println("Annotations:") for _, k := range getSortedKeys(c.Annotations) { fmt.Printf("\t%s -> %s\n", k, c.Annotations[k]) } } fmt.Println() } display.Flush() return nil } func convertContainerState(state pb.ContainerState) string { switch state { case pb.ContainerState_CONTAINER_CREATED: return "Created" case pb.ContainerState_CONTAINER_RUNNING: return "Running" case pb.ContainerState_CONTAINER_EXITED: return "Exited" case pb.ContainerState_CONTAINER_UNKNOWN: return "Unknown" default: log.Fatalf("Unknown container state %q", state) return "" } } func getContainersList(containersList []*pb.Container, opts listOptions) []*pb.Container { sort.Sort(containerByCreated(containersList)) n := len(containersList) if opts.latest { n = 1 } if opts.last > 0 { n = opts.last } n = func(a, b int) int { if a < b { return a } return b }(n, len(containersList)) return containersList[:n] }